Geoffrey Jacoby commented on PHOENIX-4195:

[~jamestaylor] - if the intention is to make it impossible to corrupt a view 
index, wouldn't you have to disable both UPSERTs and DELETEs to any 
multi-tenant table from any global connection? Even if you had a hasChildViews 
boolean in PTable, any global connection write would have a potential race 
condition if another tenant-connection was trying to create a view + view 
index, wouldn't it?

>From a usability standpoint, however, that would be problematic, because there 
>are many important use cases in which most users access a multi-tenant table 
>through tenant connections, but bulk or admin operations are done cross-tenant 
>by global connections. 

And even if we do disable writes from global connections to multi-tenant 
tables, isn't there still a race condition where another client could be 
creating a _global_ view on a non-multi-tenant table that would cause problems 
when deleting or mutating the base table?

> PHOENIX-4195 Deleting view rows with extended PKs through the base table 
> silently fails
> ---------------------------------------------------------------------------------------
>                 Key: PHOENIX-4195
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-4195
>             Project: Phoenix
>          Issue Type: Bug
>            Reporter: Thomas D'Silva
>            Assignee: Geoffrey Jacoby
>            Priority: Major
>         Attachments: test.diff
> The attached test fails.

This message was sent by Atlassian JIRA

Reply via email to